WEBVTT - San Francisco Fed President Mary Daly Talks AI, Economy 0:00:02.520 --> 0:00:07.200 Bloomberg Audio Studios, podcasts, radio news. 0:00:08.000 --> 0:00:10.520 Good morning everyone. I'd hope that we'd start by talking 0:00:10.520 --> 0:00:13.200 about productivity and some of the data that you've just said, well, 0:00:13.240 --> 0:00:18.720 no one really knows. And the question with AI and 0:00:18.760 --> 0:00:22.239 the US economy is what has happened thus far? Right? 0:00:22.760 --> 0:00:26.920 And so I'll hit you with some of the official 0:00:27.000 --> 0:00:28.760 data that I've been tracking, and you can tell me 0:00:29.200 --> 0:00:31.720 whether the utility of it or not. Right, which is 0:00:31.760 --> 0:00:34.280 really in US productivity? I always go with the Bureau 0:00:34.320 --> 0:00:38.280 of Labor Statistics measure of output per hour x non 0:00:38.360 --> 0:00:41.199 farm business sector. Right, And if you look at the 0:00:41.240 --> 0:00:43.640 data over fifty years, that chart was really interesting, the 0:00:43.680 --> 0:00:47.680 side by side of electricity and AI over fifty years, 0:00:47.720 --> 0:00:50.080 the average quarterly reading is about one point nine percent 0:00:50.320 --> 0:00:54.520 annual rate on productivity. But something's happened in the last 0:00:54.560 --> 0:00:57.160 ten quarters where it's higher, Yes, you know, just under 0:00:57.240 --> 0:01:00.360 three percent two point seven percent. Do we really know 0:01:01.000 --> 0:01:02.920 what that is? And is it AI? 0:01:03.440 --> 0:01:04.120 We don't know. 0:01:04.200 --> 0:01:06.400 I mean, that's the part that makes it hard is 0:01:06.800 --> 0:01:10.360 in productivity numbers, especially when they're happening in what I 0:01:10.360 --> 0:01:13.480 would think of as real time, it's very challenging to 0:01:13.600 --> 0:01:16.320 assess or draw it back to exactly what the factors 0:01:16.360 --> 0:01:18.280 are that have shaped it. You know, in fact, people 0:01:18.319 --> 0:01:20.320 still don't agree on what happened in the nineties all 0:01:20.319 --> 0:01:22.080 the time if you look at research. So it's just 0:01:22.120 --> 0:01:25.040 something to keep in mind. So then what you do 0:01:25.280 --> 0:01:28.600 would any good economist or person, any industry person would do, 0:01:28.680 --> 0:01:29.600 is they'd say, well. 0:01:29.520 --> 0:01:31.679 What am I seeing? What am I seeing? 0:01:31.680 --> 0:01:34.399 And so right now, while we can't find it in 0:01:34.440 --> 0:01:39.280 the macro studies, it would do very sophisticated empirical econometrics 0:01:39.319 --> 0:01:41.720 and ask the questions how much of this is AI. 0:01:42.520 --> 0:01:45.480 We still can see that there's something going on there. 0:01:45.840 --> 0:01:49.520 The question is is it happening? The question is how 0:01:49.520 --> 0:01:53.520 long will it persist? And so clearly something's happening in 0:01:53.560 --> 0:01:56.480 the economy. But if you make a series to go 0:01:56.560 --> 0:01:59.160 back to your question about productivity, if you make a 0:01:59.280 --> 0:02:02.560 series of one time adjustments, so say you automate a 0:02:03.040 --> 0:02:05.800 production line or you use AI to help in loan 0:02:05.840 --> 0:02:10.080 application process, you save money once. You don't save money forever. 0:02:10.639 --> 0:02:12.280 I mean, you keep saving that money, but you don't 0:02:12.280 --> 0:02:14.520 get growth out of that. You don't get productivity growth. 0:02:14.520 --> 0:02:17.880 You get one time adjustments to the level of productivity of. 0:02:17.840 --> 0:02:19.560 Your employees or your process. 0:02:19.960 --> 0:02:24.639 So what we're looking for is a technology to give 0:02:24.720 --> 0:02:29.880 us consistently good changes in productivity so that all industries 0:02:30.040 --> 0:02:33.200 at scale get better, industries figure out new ways to 0:02:33.280 --> 0:02:37.000 generate revenue, new ways to do product design, new ideas 0:02:37.040 --> 0:02:39.600 to come and shape the economy. That's the thing that 0:02:39.639 --> 0:02:43.519 has a sustained productivity growth part. So it's undeniable productivity 0:02:43.560 --> 0:02:46.960 growth has gone up. What's not as clear is how 0:02:46.960 --> 0:02:48.280 long will that last. 0:02:48.639 --> 0:02:53.480 Broadly, people want to see and understand how AI impacts 0:02:53.520 --> 0:02:57.600 workforce and more recently maybe inflation. So if we go 0:02:57.680 --> 0:03:01.640 back to the nineties and what green Span saw in 0:03:01.760 --> 0:03:06.400 productivity gains contributing to economic growth, there was a consideration 0:03:06.480 --> 0:03:09.079 around both of those things. Absolutely, you said that it's 0:03:09.120 --> 0:03:11.760 not the playbook to go back to what happened nineties 0:03:11.760 --> 0:03:14.320 and apply today, But what do you see in those things? 0:03:14.400 --> 0:03:18.880 Is it possible that AI is driving productivity games resulting 0:03:18.880 --> 0:03:22.560 in economic growth, but without the inflation it is. 0:03:22.520 --> 0:03:25.240 Absolutely possible and something we have to interrogate. I mean, 0:03:25.320 --> 0:03:29.960 right now, as you know to well, inflation still above 0:03:29.960 --> 0:03:34.040 our target are two percent target, and price level has 0:03:34.080 --> 0:03:36.400 been high for much higher for a long time, and 0:03:36.440 --> 0:03:40.240 people are feeling stretched by the high inflation that they see. 0:03:40.440 --> 0:03:42.960 And now oftentimes people say, well, now AI is going 0:03:43.040 --> 0:03:45.839 to take hurt the labor market, and so now I'm 0:03:45.920 --> 0:03:50.120 in double doom, as people say. But I think ultimately 0:03:50.880 --> 0:03:54.880 the way you think many people think about AI is 0:03:54.920 --> 0:04:00.560 the investment part of any technology can actually boost demand 0:04:00.680 --> 0:04:05.800 for good services in people and can then raise the 0:04:05.800 --> 0:04:09.400 pressure on inflation. But then the productivity part comes and 0:04:09.440 --> 0:04:11.760 that that's a disinflationary part. 0:04:12.080 --> 0:04:12.520 You can see. 0:04:12.560 --> 0:04:14.680 This is all about the timing, and so what we 0:04:14.880 --> 0:04:19.359 end up investigating is not just the models but asking questions. 0:04:19.440 --> 0:04:24.000 Are the buildout of data centers raising prices for construction workers, 0:04:24.320 --> 0:04:28.680 are the buildout of data centers raising prices for metals 0:04:28.720 --> 0:04:31.160 and other things that go into them? The raw materials 0:04:31.520 --> 0:04:34.560 are the productivity gains. And then on the other side 0:04:34.600 --> 0:04:37.839 of that, are the productivity gains already affecting the cost 0:04:37.880 --> 0:04:40.720 structure of firms? Do they see that and even if 0:04:40.720 --> 0:04:44.400 a series of one off adjustments can actually change the 0:04:44.440 --> 0:04:47.479 cost structure, And if you look at profit margins when 0:04:47.600 --> 0:04:50.680 prices haven't been raising as rapidly as they once were, 0:04:51.000 --> 0:04:52.919 and firms are saying they don't have as much power 0:04:52.960 --> 0:04:55.320 to pass through, you would think that they're doing something 0:04:55.360 --> 0:04:58.240 to help margin protection. And so I think this is 0:04:58.520 --> 0:05:00.840 there's something going on here. Whether we wanted to link 0:05:00.880 --> 0:05:03.160 it all back to AI and then use that as 0:05:03.200 --> 0:05:06.880 a demonstrated proof that we're in a transformative state, I 0:05:06.920 --> 0:05:11.080 think that's a little bit too far, but certainly something's happening. 0:05:11.400 --> 0:05:15.000 And thinking about the productivity growth is exactly what you 0:05:15.080 --> 0:05:17.440 know we did back in the nineteen nineties. We saw 0:05:17.600 --> 0:05:21.440 evidence firms were being more productive. We were interrogating how 0:05:21.480 --> 0:05:25.560 long that would last. And interestingly, the nineteen nineties when 0:05:25.560 --> 0:05:27.800 I said it was the Roaring nineties that followed. It 0:05:27.880 --> 0:05:30.800 was good growth, but it was also a good labor market, 0:05:31.080 --> 0:05:34.280 a really strong labor market, and so those two things 0:05:34.320 --> 0:05:37.680 went together. Because ultimately we had this conversation in the 0:05:37.760 --> 0:05:40.599 roundtable and one of the participants made a great point, 0:05:40.640 --> 0:05:41.760 it's true economics. 0:05:41.960 --> 0:05:43.239 This is how economics works. 0:05:43.320 --> 0:05:47.640 Is if an employee using AI gets much more productive, 0:05:48.240 --> 0:05:49.320 you hire more of them. 0:05:49.200 --> 0:05:50.599 Right, not fewer of them. 0:05:50.680 --> 0:05:54.320 So you know, the economy grows faster, the product development 0:05:54.360 --> 0:05:57.560 goes goes faster, and demand gets stronger. 0:05:57.600 --> 0:05:59.440 I'm going to jump ahead to data center I'd been 0:05:59.480 --> 0:06:03.359 saving it, but it's highly relevant to San Jose the 0:06:03.400 --> 0:06:06.800 build out of data center. Very recently, the CEO of 0:06:06.839 --> 0:06:09.600 PG and E, Patty Poppy, Game on the program and 0:06:09.880 --> 0:06:13.719 made the argument that it's possible that the data center 0:06:13.760 --> 0:06:16.719 build out within PG and e's jurisdiction actually brings down 0:06:16.880 --> 0:06:21.040 wholesale electricity prices because the hyperscalas take on the capital 0:06:21.080 --> 0:06:25.120 burden and they are buyers and aggregate of electricity. But 0:06:25.320 --> 0:06:28.560 many people, you know, your constituents in the twelfth district 0:06:28.839 --> 0:06:32.320 will find it hard to see that argument playing out. 0:06:32.520 --> 0:06:34.960 Well, I think we have to separate what we're talking 0:06:35.000 --> 0:06:37.279 about into now, next, later. 0:06:37.760 --> 0:06:38.880 So let's think about now. 0:06:39.279 --> 0:06:43.720 Right now, we have more demand than we have supply 0:06:44.160 --> 0:06:47.000 for energy for electricity. Right if you talk to you, 0:06:47.120 --> 0:06:50.960 we regularly have CEO round tables with the big power 0:06:50.960 --> 0:06:54.240 companies across the twelfth district. You can look throughout the nation. 0:06:54.800 --> 0:06:58.840 Demand for power is higher than the supply of power, 0:06:59.200 --> 0:07:02.080 and there's a lot of reasons why supply is falling behind. 0:07:02.440 --> 0:07:05.000 One is demand's just gone up rapidly, but another part 0:07:05.080 --> 0:07:07.320 is that they've got an aging infrastructure. They have to 0:07:07.360 --> 0:07:11.000 get those things built out. It's a highly regulated industry, 0:07:11.120 --> 0:07:13.720 so the infrastructure doesn't just come on like a light switch. 0:07:14.120 --> 0:07:16.960 Then you have there are supply chain issues that made 0:07:16.960 --> 0:07:19.840 it hard to get the transformers and other things. So 0:07:19.920 --> 0:07:23.920 all of this just adds to the imbalance of demand 0:07:24.000 --> 0:07:27.880 versus supply. But the remedy for that isn't to take 0:07:27.920 --> 0:07:30.120 away demand, it's to increase supply. 0:07:30.520 --> 0:07:31.440 So when. 0:07:33.640 --> 0:07:36.040 They talk about any CEO of a power company says, 0:07:36.080 --> 0:07:39.720 we can solve this problem by adding more supply. Absolutely, 0:07:39.800 --> 0:07:43.640 but that's a next and later. And so what you said, 0:07:43.640 --> 0:07:47.160 my constituents, what consumers and businesses are saying is I'm 0:07:47.160 --> 0:07:49.440 worried my electricity prices are going to rise, and they've 0:07:49.440 --> 0:07:54.040 already been going up. And the CEOs of power companies 0:07:54.040 --> 0:07:56.440 are saying, but if we just keep building, that will 0:07:56.440 --> 0:07:58.120 go down, and both are true. 0:07:57.960 --> 0:07:59.880 Go down as far as it will be disinflation. 0:08:00.000 --> 0:08:02.640 And you know, it's hard to say energy could be 0:08:02.680 --> 0:08:06.360 disinflationary if we get to a point where supply is 0:08:06.400 --> 0:08:08.640 greater than demand. Right now, I'm just looking for a 0:08:08.680 --> 0:08:11.080 supply to equal demand, and that would be a big 0:08:11.120 --> 0:08:13.960 benefit to consumers because it would mean that we wouldn't 0:08:14.040 --> 0:08:17.320 keep seeing inflationary pressure coming out of the energy sector. 0:08:17.680 --> 0:08:19.160 The other thing I wanted to ask you through the 0:08:19.200 --> 0:08:22.120 lens of constituents of the twelfth districts is one reason 0:08:22.640 --> 0:08:25.800 you might focus on productivity is there is a direct 0:08:25.840 --> 0:08:29.120 read through to GDP growth and other data sets that 0:08:29.160 --> 0:08:32.280 you can look at. But the anxiety in the real 0:08:32.320 --> 0:08:36.120 world is, well, a job, an AI talk can make 0:08:36.160 --> 0:08:39.640 me more productive or it can displace me altogether. Where 0:08:39.640 --> 0:08:41.880 do you see that tension in the economy right now? 0:08:42.040 --> 0:08:42.800 So one of the. 0:08:42.720 --> 0:08:46.240 Things that is true is that the labor market has slowed, 0:08:46.840 --> 0:08:49.839 but it slowed for a whole variety of reasons. And 0:08:50.000 --> 0:08:52.200 much like when you said, well, productivity is risen, Mary, 0:08:52.320 --> 0:08:55.240 so shouldn't we isn't that AI? I think we always 0:08:55.240 --> 0:08:58.800 want to be a little humble about the correlations we 0:08:58.880 --> 0:09:00.480 see and ascribing. 0:09:00.120 --> 0:09:01.080 Causality to them. 0:09:01.160 --> 0:09:04.800 So I wanted to temper your enthusiasm for thinking all 0:09:04.840 --> 0:09:07.080 the productivity growth is AI might be, but it could 0:09:07.120 --> 0:09:10.480 just be general cost management in a slowing economy or 0:09:10.520 --> 0:09:13.840 a slowing you know, or to manage tariff costs, etc. 0:09:14.400 --> 0:09:17.560 So on the labor market. The labor market is slowing. 0:09:17.920 --> 0:09:21.160 It's slowing in industries that are directly telling us that 0:09:21.160 --> 0:09:23.480 they're using AI and it's slowing in industries that aren't. 0:09:23.559 --> 0:09:25.480 So it's one of the things that I just want 0:09:25.520 --> 0:09:28.720 to be cautious. So what I talk to, We talked 0:09:28.720 --> 0:09:31.640 to workers, We talk to you know, communities all the time. 0:09:32.240 --> 0:09:36.760 What's true is in technologies is a really interesting thing. 0:09:36.880 --> 0:09:41.280 No technology ever reduces net employment, not in the history 0:09:41.320 --> 0:09:45.480 of technologies, but it does change what that employment looks like. 0:09:46.000 --> 0:09:49.920 And so there's a period of replacement right now. It's 0:09:50.000 --> 0:09:54.199 replacement of tasks. So if your job has certain tasks 0:09:54.200 --> 0:09:56.840 in it, AI can do those for you. And the 0:09:56.880 --> 0:10:01.439 next part is augmentation, so every place augment and create. 0:10:02.280 --> 0:10:06.839 What's interesting about AI is that, unlike say electricity, when 0:10:06.880 --> 0:10:10.280 the candle lighters or the lamplighters or the candle makers 0:10:10.800 --> 0:10:15.320 got displaced before the US completely became electrified, you know 0:10:15.760 --> 0:10:18.719 that this is going more quickly if you go to 0:10:18.840 --> 0:10:21.920 a firm. I was on a panel at the Reagan 0:10:22.800 --> 0:10:29.200 National Library Economic Forum with Patrick Collison from Sprint. I'm Sprints, right, Gosh, Stripe, Sorry, 0:10:29.400 --> 0:10:33.960 he's going to kill me, Stripe, don't tell him, okay, 0:10:34.440 --> 0:10:39.320 but from Stripe, and and interestingly he said, I am 0:10:39.760 --> 0:10:43.199 hiring more coders that I'm laying off, but I am 0:10:43.320 --> 0:10:47.120 laying off coders whose technology skills didn't advance or they 0:10:47.160 --> 0:10:48.080 weren't the right workers. 0:10:48.120 --> 0:10:49.440 And you're seeing this right. 0:10:49.320 --> 0:10:54.240 You're seeing you know, businesses reskill them their cells to 0:10:54.640 --> 0:10:57.359 meet the AI moment, and that's going to cause worker anxiety. 0:10:57.400 --> 0:11:00.200 And right now, worker anxiety is high. People were a 0:11:00.240 --> 0:11:04.080 low firing, low hiring environment. That's already going to make 0:11:04.080 --> 0:11:07.120 people feel vulnerable. If you haven't found a job and 0:11:07.120 --> 0:11:10.080 you're newly minted out of college, you just think I 0:11:10.120 --> 0:11:12.160 was supposed to get a job before I graduated. Now 0:11:12.160 --> 0:11:15.839 I still don't have one. That's very worriesome to people. 0:11:16.160 --> 0:11:18.880 And then if you're thinking, well, I might lose my job, 0:11:19.000 --> 0:11:20.360 but I don't know how long it will take to 0:11:20.360 --> 0:11:22.600 get another one, then you're worried about that. So I 0:11:22.640 --> 0:11:25.720 think it's natural for the sentiment to feel nervous. But 0:11:25.840 --> 0:11:29.040 it's not the same things. AI is taking all the jobs, 0:11:29.040 --> 0:11:34.200 because what we're really seeing is AIS is replacing tasks, 0:11:34.679 --> 0:11:37.160 augmenting workers. When we talk to firms, most of those 0:11:37.200 --> 0:11:39.319 firms are saying I'm augmenting my workforce. 0:11:39.600 --> 0:11:40.880 If you're in big. 0:11:40.679 --> 0:11:44.000 Manufacturing firms, they don't have enough workers that do skilled labor, 0:11:44.240 --> 0:11:47.040 and so they're looking to augment their workforce, and then 0:11:47.160 --> 0:11:50.480 we're also seeing jobs created. It's interesting I gave this talk. 0:11:50.800 --> 0:11:53.080 I gave a talk on this in twenty twenty three, 0:11:53.559 --> 0:11:56.760 and I used prompt engineers as the jobs they were creating. 0:11:56.840 --> 0:11:58.880 But those jobs are now being replaced. 0:12:00.360 --> 0:12:00.600 But a. 0:12:02.679 --> 0:12:05.439 Or it's a change, it's. 0:12:04.840 --> 0:12:07.720 A warning, you could think of it that way. Or 0:12:07.840 --> 0:12:11.080 it's an indicator. So let's take the warning. The warning 0:12:11.240 --> 0:12:15.560 is you can't keep up. I would say, let's use 0:12:15.600 --> 0:12:18.599 it as an indicator. It's an indicator that technology is 0:12:18.640 --> 0:12:24.200 evolving really fast and workforces need to skill endurable skills, 0:12:24.480 --> 0:12:25.560 and durable skills. 0:12:25.640 --> 0:12:27.280 Are be AI ready? 0:12:27.559 --> 0:12:30.360 Be able to use AI to lift yourself in the 0:12:30.480 --> 0:12:33.880 educational space. You know, use the use the technologies that 0:12:33.920 --> 0:12:36.320 are out there to build your skills up, because you 0:12:36.320 --> 0:12:40.679 can learn a lot fast if you train yourself to 0:12:40.840 --> 0:12:41.439 look at AI. 0:12:41.679 --> 0:12:44.160 Say give me a lesson on how. 0:12:44.040 --> 0:12:47.240 To write I've been thinking about this, how to write 0:12:47.280 --> 0:12:50.720 a smart contract from end to end? What sort of 0:12:50.800 --> 0:12:52.360 software would I need? How would I do it? What 0:12:52.400 --> 0:12:54.320 would the code look like? How would I test the code? 0:12:54.360 --> 0:12:56.600 How would I know it's right? Before I execute on 0:12:56.640 --> 0:12:59.720 this smart contract, and so you can do these things 0:13:00.160 --> 0:13:03.000 in an evening and then it's just about being able 0:13:03.040 --> 0:13:05.720 to do that. So I think that's the message for workers, 0:13:06.120 --> 0:13:08.480 and I would have taught my young self this same thing. 0:13:08.679 --> 0:13:14.600 Is if you put off technology because you're afraid of it, 0:13:15.320 --> 0:13:19.319 then you won't be in the first place of trying 0:13:19.360 --> 0:13:22.520 to use the technology to further your own abilities. 0:13:22.600 --> 0:13:25.240 Can we extend that to the FED? Now bear with 0:13:25.280 --> 0:13:27.920 me on that one. Okay, all right, you talked about 0:13:28.000 --> 0:13:34.120 disaggregated data but also improved measurement, citing Greenspan in that sense. 0:13:34.120 --> 0:13:39.959 If AI is so good, can it process larger sets 0:13:40.000 --> 0:13:45.000 of data and make more accurate economic forecasts than traditional 0:13:45.000 --> 0:13:45.640 FED models? 0:13:45.679 --> 0:13:50.600 Can? We know we don't right now use AI in 0:13:50.600 --> 0:13:55.120 our monetary policy work, but we do use it in 0:13:55.200 --> 0:13:58.360 research as researchers. If you go to any academic institution, 0:13:58.800 --> 0:14:03.280 you're going to see researchers using AI to see what 0:14:03.360 --> 0:14:05.440 they can do better on coding and other things, but 0:14:05.480 --> 0:14:07.319 also data analytics. 0:14:07.480 --> 0:14:10.719 What do you see The place. 0:14:10.480 --> 0:14:14.760 We are there is AI doesn't give you answers to problems. 0:14:14.800 --> 0:14:17.920 It helps you get to the discovery perspective. So if 0:14:17.960 --> 0:14:20.840 I use AI as a researcher to look at a 0:14:20.880 --> 0:14:24.160 bunch of data. I still have to test my hypothesis. 0:14:24.160 --> 0:14:26.080 I have to go in with a hypothesis. What am 0:14:26.120 --> 0:14:28.680 I trying to answer? So that's the human person. And 0:14:28.720 --> 0:14:32.480 so that's why it's not particularly well tooled right now 0:14:32.760 --> 0:14:39.840 to replace our forecasters and our thinkers, our scholars who. 0:14:39.800 --> 0:14:41.920 Comproduce a more accurate neutral rates. 0:14:41.960 --> 0:14:45.080 For example, Well, you're still going to get estimates that 0:14:45.120 --> 0:14:47.920 are between eleven and negative three on the neutral rate 0:14:47.960 --> 0:14:50.960 of interest, So I'm not kidding. Models can the models 0:14:51.040 --> 0:14:54.360 that we have can produce an estimate from negative three 0:14:54.720 --> 0:14:58.320 to positive eleven, right, And so there. 0:14:58.320 --> 0:14:59.240 What does that tell you? 0:14:59.640 --> 0:15:02.440 The the neutral rate of interest is not a truth 0:15:02.520 --> 0:15:03.320 with a capital T. 0:15:03.840 --> 0:15:04.680 It's an estimate. 0:15:04.720 --> 0:15:07.720 It's a theoretical construct to help us understand how to 0:15:07.760 --> 0:15:11.920 benchmark policy. But you can't use it as a threshold 0:15:11.960 --> 0:15:15.560 that you can do surgical adjustments around. No one calibrates 0:15:15.840 --> 0:15:20.040 monetary policy surgically with a neutral rate of interest estimate 0:15:20.160 --> 0:15:23.360 for those reasons. So we are using AI though at 0:15:23.360 --> 0:15:25.480 the FED, and many people may be surprised about that. 0:15:25.520 --> 0:15:26.640 Would you like to learn about that? 0:15:26.960 --> 0:15:27.840 Yes? Please help? 0:15:28.920 --> 0:15:32.160 So I know many might think I work in an 0:15:32.160 --> 0:15:35.800 institution that waits for elect we're still getting electricity. You 0:15:35.840 --> 0:15:39.520 might think that, but no, we actually are not the 0:15:39.560 --> 0:15:43.480 earliest adopters because remember we're fiduciary stewarts of public funds 0:15:43.560 --> 0:15:46.480 but also as fiduciary stewards of public trust, and so 0:15:46.520 --> 0:15:48.720 we really have to make sure that we're working in 0:15:49.120 --> 0:15:53.560 the most risk free and risk managed environment we possibly can't. 0:15:54.520 --> 0:15:57.800 But we have been at this since really in twenty 0:15:57.840 --> 0:16:00.120 twenty three. So the first thing that we did it 0:16:00.160 --> 0:16:02.200 as a system, and I'll really speak about the twelve 0:16:02.320 --> 0:16:04.920 at a reserve banks that are across the country. We 0:16:05.000 --> 0:16:08.040 worked as a system to say, well, we need to 0:16:08.080 --> 0:16:13.520 make sure our employees, our teams are ready to understand AI. 0:16:13.800 --> 0:16:15.240 So what do we need to do. 0:16:15.320 --> 0:16:19.760 We need to have lessons, work playing, you know, work gatherings, 0:16:19.760 --> 0:16:23.520 et cetera, get people familiar and get people focused in 0:16:23.640 --> 0:16:25.800 areas that we can practice with. So we built a 0:16:25.840 --> 0:16:30.080 practice environment that was completely ring fenced and not in production. Right, 0:16:30.120 --> 0:16:32.480 it's just a practice environment trying it out, and of 0:16:32.520 --> 0:16:35.000 course we got what most businesses got. The other businesses 0:16:35.040 --> 0:16:36.280 did exactly the same thing. 0:16:36.480 --> 0:16:38.239 And what do you get? You get the early adopters. 0:16:38.400 --> 0:16:40.200 But the good news about our early adopters as are 0:16:40.200 --> 0:16:43.440 often ambassadors. So then we're holding like tech cafes and 0:16:43.480 --> 0:16:45.280 things to help other people learn that. 0:16:45.400 --> 0:16:46.080 So that was then. 0:16:46.400 --> 0:16:48.880 So then in twenty twenty four and twenty five we 0:16:48.960 --> 0:16:52.200 really made a full court press push March Madness is 0:16:52.200 --> 0:16:52.640 coming up. 0:16:53.720 --> 0:16:53.920 You know. 0:16:53.960 --> 0:16:57.280 We really went hard at making sure that people had 0:16:57.320 --> 0:16:59.840 not just the if they're interested, do it, but that 0:17:00.040 --> 0:17:01.400 this is something that we really want. 0:17:01.320 --> 0:17:01.640 You to learn. 0:17:01.720 --> 0:17:03.640 This is the operations within the system. 0:17:03.680 --> 0:17:05.639 This is the operations. I should have said that before. 0:17:05.920 --> 0:17:08.000 I'm sorry. Let'll tell you that if you were at 0:17:08.000 --> 0:17:10.720 a reserve bank, and again little known facts. These are 0:17:10.720 --> 0:17:14.160 like facts that people don't know about the FED if 0:17:14.160 --> 0:17:16.840 you go to a reserve bank or any of our operations. 0:17:17.080 --> 0:17:20.080 Most of the people who work with us are operations people. 0:17:20.320 --> 0:17:26.159 We process cash, We do all the electronic payment system backbones, 0:17:26.280 --> 0:17:28.119 make sure they operate on time. If you're in the 0:17:28.119 --> 0:17:31.159 financial sector, you know FED wire or acch FED Now 0:17:31.400 --> 0:17:32.840 all of that is operated by. 0:17:32.760 --> 0:17:35.760 Our operations teams. We also support Vice Chair. 0:17:35.720 --> 0:17:39.960 Bowman in supervision of banks and all of those things. 0:17:40.800 --> 0:17:42.600 And then we have all our support people who help 0:17:42.640 --> 0:17:45.560 make sure that that occurs. All of that can be 0:17:45.640 --> 0:17:48.360 easily if you can do AI and you can use it, 0:17:48.560 --> 0:17:51.840 you can think of opportunities. So the next thing we did, 0:17:51.960 --> 0:17:54.720 get our workforce ready is number one. For next thing 0:17:54.760 --> 0:17:57.080 you do is and this is again like all the 0:17:57.119 --> 0:17:59.800 businesses we talk to, you see what your vendors are 0:17:59.800 --> 0:18:02.720 our offering, right if you're you have a technology vendor, 0:18:02.760 --> 0:18:05.320 an accounting vendor, an HR vendor, and then you just 0:18:05.359 --> 0:18:07.560 turn the service on. But if you turn the service 0:18:07.600 --> 0:18:10.160 on before your team is ready, you don't really get 0:18:10.160 --> 0:18:12.560 an ROI out of it. So again, fiduciary stewarts of 0:18:12.600 --> 0:18:15.360 public funds, we have to make sure so we're all 0:18:15.480 --> 0:18:19.440 instilled in this ring fenced proprietary environment because the public 0:18:19.480 --> 0:18:22.520 has to know that we're not introducing risk to this. 0:18:23.040 --> 0:18:25.440 And then of course the place we're. 0:18:25.280 --> 0:18:28.320 Working now is where many many people are working, which 0:18:28.400 --> 0:18:31.760 is if you have a lot of technology workers, then 0:18:31.880 --> 0:18:36.760 the coding assist is just so important and we're one 0:18:36.760 --> 0:18:42.280 of the things back to workforce. It doesn't create a 0:18:42.400 --> 0:18:45.199 massive change in who you have working for you. What 0:18:45.240 --> 0:18:47.879 it means is they can do their work faster, better 0:18:48.160 --> 0:18:50.520 and more effectively. And if you think of the three 0:18:50.800 --> 0:18:55.840 timples of the FED, we need to be efficient, effective. 0:18:55.480 --> 0:18:56.280 And resilient. 0:18:56.480 --> 0:18:59.439 So it also builds in that resilience for us because 0:18:59.440 --> 0:19:03.560 we have you know, quality assurance and unit testing. All 0:19:03.600 --> 0:19:05.240 the things that can slow you down if you're not 0:19:05.359 --> 0:19:09.840 right or interrupture your ability to serve. Those things can 0:19:09.880 --> 0:19:13.080 all be assisted with AI in a really positive way. So, 0:19:13.600 --> 0:19:18.960 again not monetary policy, but definitely like all other companies 0:19:19.240 --> 0:19:22.000 who are working on this space, making sure we're not 0:19:22.119 --> 0:19:25.840 behind and delivering good value for I mean, the shareholders 0:19:25.920 --> 0:19:29.000 of the FED are the American people, and we owe 0:19:29.040 --> 0:19:32.919 them the effort to make sure we're modernizing ourselves and 0:19:33.000 --> 0:19:35.080 keeping up with the things that can help us do 0:19:35.119 --> 0:19:36.920 our work faster, better, and more effective. 0:19:37.080 --> 0:19:40.480 I do have two questions relating to AI and monetary 0:19:40.480 --> 0:19:42.520 policy quickly, and I know we want to get some 0:19:42.600 --> 0:19:44.720 audience questions as well, some conscious there are students in 0:19:44.720 --> 0:19:47.480 the room who will go out into the workforce. I 0:19:47.560 --> 0:19:50.280 think that the main thing reflecting back on the nineties 0:19:50.880 --> 0:19:55.320 is that there are anticipated impacts yes from AI on 0:19:55.359 --> 0:20:01.760 the economy and PCEE is the preferred gauge inflation running 0:20:02.200 --> 0:20:06.080 higher beyond two percent. How do you manage that. You know, 0:20:06.119 --> 0:20:10.880 many would argue that those anticipated AI driven productivity gains 0:20:11.160 --> 0:20:15.639 would justify lower rates, but they are that anticipated, yeah, and. 0:20:15.640 --> 0:20:18.480 I think that, you know, really, it's important to recognize 0:20:18.640 --> 0:20:21.439 that monetary policy is a forward looking business, but it's 0:20:21.480 --> 0:20:24.240 also an evidence based business. And so there will be 0:20:24.320 --> 0:20:28.439 a point in time when we'll have enough confidence that 0:20:28.480 --> 0:20:33.240 the anticipated effects are materializing. And where would you look 0:20:33.280 --> 0:20:35.800 for that? You look for that in what's happening with 0:20:35.840 --> 0:20:39.280 price pressures? Not just aggregate inflation, but if you disaggregate 0:20:39.320 --> 0:20:41.840 it and you ask yourself a question, are the AI 0:20:42.240 --> 0:20:46.720 using sectors just doing less pass through of input costs 0:20:46.880 --> 0:20:49.720 into prices? Well, maybe that tells you something. So that's 0:20:49.720 --> 0:20:51.480 where the research really becomes important. 0:20:51.560 --> 0:20:52.360 You ask questions. 0:20:52.359 --> 0:20:56.880 But you also do research where you can disaggregate firms, 0:20:56.880 --> 0:20:59.159 you can disaggregate prices, and you can ask where do 0:20:59.160 --> 0:21:00.800 we see price pressure. 0:21:00.680 --> 0:21:02.719 And how do we think they will evolve. 0:21:03.440 --> 0:21:06.200 That's so that's important, and you can't wait because remember 0:21:06.240 --> 0:21:08.960 Monterrey policy as a twelve to eighteen month lag So 0:21:09.080 --> 0:21:13.480